The Rise of AI Image Generation Models and Comparisons

AI image generation is rapidly evolving, with various models excelling in realism, text incorporation, and unique creativity.

### The Evolving Landscape of AI Image Generation: Innovations and Comparisons
* 🖼️  AI image generation has recently experienced a **significant surge in innovation**, with models like Grock 2 producing **increasingly realistic and diverse images**, though some content may face restrictions.
* 🔍  Grock 2 operates on the Flux One model from Black Forest Labs, which has been a focus in recent videos showcasing techniques for **creating realistic images** and **training personal models**.
* ✨  A new contender, idiogram 2.0, has emerged as a **free text-to-image generator** that excels at **incorporating text into images**, despite lacking some advanced features found in other models.
* 🧪  In a comparative analysis, various prompts were tested across multiple AI models to evaluate their performance in **human realism**, **landscape depiction**, **text integration**, and **absurdity**.
* 🌅  Results showed idiogram 2.0 generating **impressive images** for a serene Japanese Zen garden, while Mid Journey produced **high-quality images** but struggled with **text adherence** in complex prompts.
* 🤖  The Freepik Mystic model and Leonardo's Phoenix also demonstrated solid capabilities, particularly in **realism and aesthetics**, with Phoenix gaining praise for its **color contrast**.
* 💰  As competition intensifies, users can access various free options, including **10 daily images from Idiogram** and **25 images from Mid Journey's trial**, enhancing the choices available for AI image generation.
* 🎨  Ultimately, the rapid evolution of AI image generation tools offers consumers a **plethora of options tailored to different needs**, from realism to text integration, making it an **exciting time for creators**.

AI image generation has seen a huge surge in innovation lately. By now, we've probably all seen the crazy, uncensored, unhinged images coming out of Grock 2. They're realistic; they'll generate pretty much anything, but like nudie content, and they're probably going to end up getting excluded from all sorts of different directions. However, under the hood, Grock 2 is just using the Flux One model from Black Forest Labs. And if you've been watching this channel at all lately, you'll have noticed that I've been pretty much going ham on using Flux lately. My latest videos, some of the more popular videos around the topic, were this one where I showed you how to make super realistic images, and this one where I showed you how to train your own face into the model. Well, this week, we've gotten even more AI image generation rollouts, and quite honestly, it's starting to feel like Mid Journey is taking notice and making some changes to make their product a little bit more appealing to you. But let's just start

with what was released most recently this week, we got a brand new mod model from idiogram called idiogram 2.0. It's their most advanced text-to-image model, and it's now available to all users for free. Now, if you're not familiar with idiogram, I made some videos on it several months ago, but it was pretty much the first text-to-image generator that was really good at actually adding text into your images. And I actually believe it's using its own Foundation models, so it's not built on top of something like stable diffusion or Dolly 3 or flux or something like that. It's its own unique model. And with this latest rollout of idiogram 2.0, it's pretty much on par with most of the other AI image generators out there. The problem is, since it's not an open model, we don't have things like control Nets and luras and IP adapters and inpainting and outpainting and all of the stuff that you can do with all these other models yet. However, one thing it does

have going for it is that anyone can use it for free right now. So in this video, what I want to do is I want to put idiogram to the test. I want to play with some other new AI models that are coming out, and I want to sort of pit them against each other and see what's good at what. So in order to do this, I actually used Claude to generate four different prompts that I'm going to use in a bunch of different image generators, and we'll see how they all stack up against each other. I'm going to be testing for four things in this video. There are a lot more things I can test for, and I'll probably do more tests like this in future videos, but for this video, I'm going to test human realism, landscape and scenery, incorporating text into images, and just weird and absurd images. And I had Claude generate prompts for each one of those. So for example, for human realism, I used the prompt a close-up portrait of a weathered elderly fisherman with deep wrinkles wearing a

yellow raincoat and knit cap against a stormy sea background. And when I plugged that into idiogram, I got this image, this image, this image, and this image. They all look pretty decent. I mean, this one looks like he might be blind or something. This one to me is looking a little bit noisy, same with this one. I'm seeing a little bit of noise, but this one to me is really, really impressive. Next up, I wanted to test landscape and scenery, so I did a Serene Japanese Zen Garden at Twilight with carefully raked sand patterns, moss-covered rocks, and cherry blossom trees in full bloom. Here's the first image the idiogram gave me. Here was the second. Here was the third. And here's the fourth. And they all look pretty good, no complaints with any of them. So incorporating text, this is idiogram has and still does really stand out. So I did a mystical forest clearing where wisps of fog spell out the words magic awaits in flowing ethereal lettering between ancient trees. Here's generation number one. You can see

looking pretty good there generation number two. I really love the colors of this one. However, the background looks a little wonky. Here's the third generation. Looks pretty good. All of the text is perfect in all four of these images. And here is the four one. Magic awaits and it looks excellent. And finally, for weird and absurd, I did a steampunk-inspired octopus riding a unicycle made of clockwork gears juggling neon cubes while floating in bubble tea. So, a lot of elements. This is kind of testing weirdness and absurdity, but also testing how much of the things it actually gets into the image. How adherent is it to my prompt? And well, here was the first one that it came up with. We've got our octopus. We've got our unicycle made of clockwork gears juggling neon cubes while floating in a boba tea. Now, it kind of put the Boba T on the side, so it's not really in the sea, but it did a pretty damn good job of getting all of those elements in there. Here's the

second generation. We've got our cubes. These kind of look more like marbles to me. We've got our Clockwork gear unicycle. Once again, it pretty much got all the elements. Here's the third one. Not seeing the Boba T element, but everything else it did a pretty good job on. And here's number four. Once again, got everything. We see the steampunk style, the cubes. Hard to say this is Boba tea, but pretty dang close with all the elements. IDog 2 is pretty impressive. Now on their tweet here, they say it is free for all users. There is a caveat. If we click on their manage subscription page, we can see that you can get 10 credits per day. Since it generates four images per credit, that'll generate about 40 images per day. Beyond that, you do have to pay, but you can do 10 generations of four images each every single day with it. Now with Flux and Idiogram getting all of this attention lately, I think Mid Journey is starting to take notice. They're probably starting

to feel a little bit of heat because all of these free and just-as-good alternatives are popping up, and many of them are either free or much less expensive to use. Now, I don't know if this is a coincidence or not, but the same day that idog announces 2.0 and that you can actually use it for free, Mid Journey made this announcement: The Mid Journey web experience is now open to everyone. We've also temporarily turned on free trials to let you check it out. So, idiogram drops the day that I'm recording this, Mid Journey announces they're opening up free trials the day that I'm recording this. So now anyone can generate a handful of images inside of Mid Journey. Now, the free trial with Mid Journey also comes with a caveat. If we check out their announcement in the Discord here, they say, Hey everyone, today we're opening up the web image creation to everyone to let people play with the new site. We've temporarily turned on free trials, roughly 25 images, so you get a total of 25 images to test it, and I

think that's total. I think once you've gone through your 25 images, you're done. You got to pay for Mid Journey to use it anymore. That's not per month, per day, anything like that. I think 25 images. If you want to keep using Mid Journey, you got to pay. Unfortunately, Mid Journey hasn't rolled out a ton of new amazing features lately. They did sort of update the way they do in painting, and they've added some cool features over the last few months, but nothing super recently. But of course, I wanted to test the same prompts inside of Mid Journey to see how they compare to what we got out of Idiogram. And don't worry, you don't need to rewind the video to see what I did on Idiogram. I'll show you the side-by-side comparison a little bit later in the video here. But when I did a close-up portrait of a weathered elderly fisherman with deep wrinkles wearing a yellow raincoat and knit cap against a stormy sea background, it actually generated four pretty dang decent images.

they're all pretty realistic. I almost feel like the wrinkles are almost too HD to feel like perfect, but they're really, really good images. Here's our serene Japanese garden at twilight with carefully raked sand patterns, moss-covered rocks, and cherry blossom trees in full bloom. It looks like it got all of those elements. We've got the moss-covered rocks, the sand patterns, the cherry blossoms, twilight. It hit on all the marks. So, Midjourney is pretty much doing it just as good as Ideogram text is, where Midjourney kind of falls apart a little bit right now. So, a mystical forest clearing where wisps of fog spell out the word magic awaits in flowing ethereal lettering between ancient trees. Well, the trees and the fog look amazing, but none of this looks like magic awaits to me. This looks like Migy aists and that's Maggie and imp and then Maggie gigus, something like that. And then finally, a steam-inspired octopus riding a unicycle made of clockwork gears juggling neon cubes while floating in a

bubble TC yeah mid Journey just kind of falls apart when you want it to adhere to a whole bunch of things in a single prompt. I don't see anything that says steampunk about these. None of these look like unicycles. We've got the octopus, we've got the cubes, but the rest kind of didn't make the cut for this generation. But like I said, there's been a lot happening in the world of AI image generation, and well, just talking about idiogram in mid journey is not really a lot. We got another new AI image generation model from the company Freepik. Now, this company Freepik, it's kind of similar to Canva, but they recently acquired the AI upscale platform Magnific. And this new Mystic model seems to be kind of the first real collaborative thing they've done together since the companies have combined forces. Now, personally, I'm unclear if this is a brand new foundation model that they created or if they, you know, fine-tuned or added their own sort of pipeline on top of like a

stable diffusion or a flux or something like that. I'm not totally sure. I'm sort of operating under the impression that it's its own foundation model, but I can't conclusively say that right now. However, just like the other alternatives, the outputs are pretty dang good. Now, this one only generates one image at a time when I give it a prompt, but here's a close-up portrait of a weathered elderly fisherman with deep wrinkles, wearing a yellow raincoat and knit cap against a stormy sea background. It got it all, except for the sea. It's not really super clear behind it. A serene Japanese Zen garden at twilight with carefully raked sand patterns, moss-covered rocks, and cherry blossom trees in full bloom. It pretty much got all of that, except I'm not really seeing the sand rake pattern on this one. It also kind of struggled with the text. It's pretty close, but you can see that this doesn't really look like a word. It sort of blended the W and the A together on a weights, not a

horrible image by any means. It just sort of fell apart a little bit on the text. And then finally, here's our octopus. I'm not seeing anything about it that tells me this is a steampunk octopus. It did seem to get the Clockwork gears a little bit. I mean, it's got a clock. I don't know if I'd really consider that gears, but it's got a clock, and it's got the neon cubes and the bubble TC. I could. I can buy it. These look like little bubble te bubbles. I mean, they're pink, but I'll buy it. Now, similar to idiogram, free pick has a limited amount. They have daily limits to AI image generation. Here, I'm not actually sure what the limit is at the exact moment, because if we take a look at this Mystic model, you can see that it's still in Alpha, and it actually says coming soon. They actually sent me early access, so I can play around with this a little bit ahead of time and show off what it's capable of. But it's going to

be rolling out really, really soon, and we'll know more about what sort of limits they're going to put on you once it's actually publicly rolled out. And of course, we still have the amazing model from Leonardo called Phoenix, which in my opinion generates some pretty dang amazing images as well. Well, here's what I got with my old man in the sea prompt down here. It seemed to have gotten everything I asked for in these prompts, and quite honestly, I'm pretty impressed with the images that come out of it. I love the contrast of what Leonardo Phoenix gives you. Here's our zen garden with the moss-covered rocks and all that stuff from The Prompt. Did it perfectly. It also did the text pretty good on three out of four. So, we got magic awaits on this one, this one, and this one. They all look great. This one in the middle a little funky with the text, but it actually does a pretty good job on the text. And once again, I'm loving the colors. I just, I think Phoenix does really, really great colors. And then

finally our octopus here. I'm going to go ahead and pick this one and say it's probably looking the best. We've got our steampunk elements, we've got our Clockwork gear unicycle, we've got it juggling cubes, and we got our bubble tea. I mean, they're like bubbles, close enough, looks pretty good. Now, I am an adviser to Leonardo, mostly because I think it's a fantastic product, but I'm always a little worried that I'm going to come off too overly biased towards them because I am an adviser. But I do genuinely think it is as good as any of the other models that are out there right now. And of course, you still have all the originals that are available out there. You've got Stable Diffusion, you've got Dolly 3, you've got Adobe Firefly, we've now got Imagen 3 from Google. There are so many other image generators out there, and to be quite honest, they've all kind of caught up to each other. In fact, I made this board here inside of

figma that compares all of the models that I'm aware of and have access to at the moment to play with. And you can see I put the four prompts across the left side here and then all of the various models across the top here. So I've got idiogram 2.0, mid Journey 6.1, free pick, Mystic, Leonardo's Phoenix flux .1, which I generated directly inside of grock. We've got Dolly 3, stable diffusion 3, Firefly 3 from Adobe, meta emu image in 3 from Google, and then we've got playground V3 over here. I'll actually link up to this figma board in the description below if you want to take a peek because one thing that's really cool about figma, and I've only recently started playing with figma, is if I hold down control and then zoom in, all of these images are actually full quality when we zoom in on them. So you can actually look at some of the more fine details of any one of these images that

I've shown off by zooming in on this so you can see here when we look at idiogram. Here's our close-up portrait of a weathered elderly fisherman with deep wrinkles, etc., etc. Here's idiogram. I'll zoom in on this one, move over here. You can see mid Journey 6.1 compared to it. Slide over a little bit. Here is the free pick Mystic version of it. Here's the Leonardo Phoenix version of it. Flux one that I generated in Gro. Here's the dolly 3 version that I generated. Stable diffusion 3. Firefly 3, which kind of looks like Santa wearing a yellow jacket, honestly. Here's the meta emu version of it. The imin 3 version of it. And the playground version 3 of it. All of them did a pretty dang good job. And I think what we're seeing here is that pretty much all of the models have kind of converged and are pretty much equally as good at realism. If there's one that I

say is a low point for realism. It's probably this one here with Firefly 3. It just looks a little off to me here, and then this one here from Playground 3. It would be decent, but it made the eye just a little too blue. With our Zen garden, here's how Idiogram compares to Mid Journey 3 right here. Here's how it compares to Mystic. Again, you can see that it's sort of missing the patterns in the sand, but other than that, it got pretty much everything. Here's the Leonardo Phoenix version, the Flux version, the Dolly 3 version, our Stable Diffusion 3 version, the one that came out of Firefly, and then we have Emu Imagin and Playground 3 for that prompt. So, you can kind of see the comparison, and again, I will link this in so you can zoom in and scroll around and look at it all. Now, taking a peek at incorporating text once again, here's our image in two model versus Mid Journey, which doesn't quite do the text very well. Here's the

free pick Mystic version Leonardo Phoenix flux one Dolly 3, which I'm really impressed with how good it's doing text these days. It wasn't doing text very well a couple of months ago. Here's stable diffusion 3, kind of didn't get the text amazing, but for the most part, pretty good. It's just messed up the G. It kind of looks more like a backwards R. Here's Firefly. It just absolutely refuses to do text in the image. Didn't even try. Here's meta's emu, actually pretty good at text. Imagine 2, pretty good at text again. And then here's the playground V3. It actually did a decent job of the text. It actually kind of more looks like the text is made out of smoke, so pretty impressive on that one. And then finally, when we're looking at our prompt aderit here on our weird abstract steampunk octopus image in 3, kind of crushed it. Here's mid Journey. We already talked about that one. It's kind of missing a few of the elements. Here's the

version out of free pick Mystic. Our version out of Leonardo Phoenix. Here's what Flux One generated. It kind of put human legs. It didn't do a unicycle. It kind of made it more like a bicycle. They're definitely not in a sea of boba tea, but it's got an octopus in some cubes, but no clockwork, no real steampunk. Yeah, Flux did not do great with the prompt adherence. Here we have Dolly 3. Dolly 3 also kind of crushed it. We've got the steampunk, we've got the clockwork unicycle, the cubes, the boba. It got it all in there. Dolly 3 to this day still hands down the best at being prompt adherent. Here, Stable Diffusion 3. It got some of the elements. There's no steampunk here. The unicycle looks all jacked up. It doesn't look like a clockwork unicycle. There's nothing to do with bubble tea, boba tea here, but it did get cubes in an octopus in there and

I actually do kind of like the colors of it. I think it did a good job with the color palette. Here's what Adobe Firefly generated. Quite honestly, better than I would have expected out of Firefly. We've got the steampunk. We've got a clockwork-looking unicycle. We've got some bubbles. We've got the octopus. There's no cubes, nothing that says this is like bubble tea, but it got a lot of the elements there. Pretty surprising, honestly. Here's Meta's emu. Also put legs on more of like a bicycle, motorcycle-looking thing, but it does have the steampunk aesthetic, the octopus, and the cubes. Here's Imagine 3. Octopus, cubes, unicycle, steampunk legs, boba. Most of the elements are there. And then finally, here's Playground V3. We got a big old jug of boba here. We've got some cubes. We've got our octopus. We've got our unicycle. There's some gears here. I'm not seeing the steampunk aesthetic, but it got quite a bit of the

elements not my favorite sort of color palette and aesthetics, but it did a pretty decent job. So, all this to say, pretty much all these image models are almost just as good as each other in most of these areas. I do want to do some deeper dive testing. You can see here I put more prompts to come because I want to think of more prompts to test different areas of these models, and then I'll keep on building this out and adding new images with new prompts to test different types of outputs that I want to try to get from these models and see how they do. But if you're sitting there going, I want to get into AI image generation. I don't know what model to use, hopefully, this helps you sort of gauge what's better at what. I'd say Dolly 3 is still the king at prompt adherence. So, if you have like nine things in your prompt and you want to make sure they're all in the image, Dolly 3 is going to crush that for you if you

want realism. They're all pretty much catching up to each other. I think Flux One is probably the king of realism at the moment. Mid Journey is pretty good. I actually don't think this Mid Journey image does Mid Journey justice on realism. I've seen a lot better, more realistic images than what I got out of this specific prompt. But Mid Journey is pretty dang good at realism. I mean, Idiograms looking pretty dang good at realism. You really can't go wrong. Firefly 3 and Playground V3 are the two that I'm like, N, I don't know about that. If you want to get text in your image, Idiogram is definitely going to be the king of that one right now. But Phoenix does an amazing job. Lux seems to do a pretty good job. Dolly 3 does a really good job. Imagine 3 is doing a pretty good job. Playground V3 is doing a pretty good job. Most of the models are doing a pretty decent job at getting the text into the image. So maybe just avoid

Firefly, if you need to get that text in it and maybe avoid mid Journey if you need to get text in the image, mid Journey is not doing amazing at that still. And if you're just looking at overall aesthetics, that's just a little too subjective for me to tell you what's the winner. We're all going to have different tastes on what we think the best aesthetics are in an image. Personally, I know I'm biased, but personally, I love, love the aesthetics of Leonardo Phoenix, and I also think typically mid Journey does a really amazing job with the aesthetics and the color contrast and things like that. But again, I'm going to do more tests. I'll probably make a part two, add four more prompts to this, and test different areas that I didn't test in this video. I will link up this board for Fig Mut down below so you can get in, zoom in, look at the fine details, see if there's anything that you don't like or do like about them, and maybe help you.

make a decision on what the best image generator is for your needs. Again, let's go over pricing real quick. Idiogram 2.0 is free for 10 images a day. Mid Journey 6.1 offers 25 images total on the free trial. Mystic is not quite available yet. Leonardo Phoenix has a limited amount of free credits. Flux one: if you are a grock user, you can use it directly inside of grock. But I also made a video about how to use it with glyph, where you can actually generate images with glyph. So check that out. Dolly 3: you can generate 100 images a day for free using Bing's image Creator. Stable diffusion 3: I'm sure there's some free models out there. I think you can find it on hugging face and generate on hugging face. Firefly 3: I think you need to be an Adobe Creative Cloud member. I'm not 100% sure. I'm an Adobe Creative Cloud member, so I was able to do it within my Creative Cloud. I don't actually know the pricing to just

use Firefly. It may just be free. Meta's emu. You can use it for free directly inside of Instagram and WhatsApp and Facebook Messenger. Imagine 3. You can use it inside of Google's AI Test Kitchen for free right now. Playground version three. They give you a certain amount of credits per day to use it for free right now. The most expensive model you can possibly use is Mid Journey, and Mid Journey knows that, which is why they just opened up their free trial again. And that's my long, probably over-complicated breakdown of the current landscape of AI image generation. But the beautiful part about all of this is that as this competition heats up, us as the consumers, we're the ones who win. We have free options. We have super realistic options. We have options that'll do text. We have uncensored options. We have ultra-prompt adhered options and pretty much everything in between. If you have an idea in your brain for an image, there is a tool out there right now that will be able to generate that image. And to me

that's pretty cool. Hopefully, you enjoyed this video. If you do, give it a thumbs up. Maybe consider subscribing to this channel if you want to hear more about AI tools and news and the latest announcements and cool stuff in the AI world. I'd really, really appreciate it. Check out future tools. That's the site where I show off all the cool AI tools that I come across, and I really appreciate you nerding out with me over AI images today. I'll see you in the next one. Bye-bye.

Midjourney is Free Again... and here's why.

AI, No-Code, Tech, Futurism - I'm a tech nerd and talk about tech nerd stuff

Matt Wolfe

AI image generation has rapidly evolved, with Grock 2's realistic and uncensored images causing a stir.

The model behind Grock 2, Flux One from Black Forest Labs, has been a focus of recent videos.

0 was released, offering a unique text-to-image generation model for free, aiming to compete with other AI image generators.

The second generation of images showed great color and perfect text, while the background needed improvement.

The steampunk-inspired octopus prompt successfully generated images with all the elements, though the boba tea placement was off in some.

0 offers 10 free credits per day for image generation, while Mid Journey has also launched free trials, offering 25 images to test before requiring payment.

AI Image Generation: Midjourney and Freepik's Mystic Model

Midjourney's image generation is impressive but struggles with complex prompts, resulting in inconsistent outputs.

The company Freepik, after acquiring Magnific, has launched its Mystic model for AI image generation, but it's unclear if it's a new foundation model.

The Mystic model produces good outputs, generating one image at a time, but also struggles with certain elements, like clear backgrounds and text.

The steampunk octopus image has clock elements and neon cubes, but its connection to steampunk is weak.

Free Pick offers limited daily AI image generation, with a public rollout of their Mystic model coming soon.

Leonardo's Phoenix model impresses with its image quality and color contrast, while also offering a variety of image generation options.

The Figma board showcases a comparison of various AI image generation models, including Idiogram 2.

Each model's output is displayed for four different prompts, allowing for a detailed comparison.

Overall, the models have converged in their ability to generate realistic images, with Firefly 3 and Playground V3 being the least effective in realism.

Comparing AI Image Generators' Text and Prompt Adherence

Mystic's version and Leonardo Phoenix's Flux One have recently improved in text generation, while Stable Diffusion 3 and Firefly struggle with it.

Dolly 3 stands out as the best in prompt adherence, successfully capturing all elements of the steampunk octopus image.

Other generators like Meta's emu, Imagine 2, and Playground V3 also perform well, but none fully achieve the desired steampunk aesthetic.

The image models being tested are all performing similarly well in various areas.

Dolly 3 excels at prompt adherence, while Flux One is currently the best for realism.

Idiogram is leading in text integration, and overall aesthetics are subjective, with Leonardo Phoenix and Mid Journey being personal favorites.

0 offers 10 free images a day, while Mid Journey 6.

Dolly 3 allows for 100 free images a day through Bing's Image Creator, and Meta's emu is free to use on Instagram, WhatsApp, and Facebook Messenger.

The competition among these tools is increasing, providing consumers with a variety of free and realistic image generation options.